In a logistic regression a positive/negative beta tells you that the direction that variable works. Is there anything in the Random Forest variable importance measures that indicates the direction of the variable?
1
-
$\begingroup$ github.com/dchudz/misc/wiki/Average-predictive-comparisons $\endgroup$– assumednormalCommented Nov 27, 2014 at 2:33
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
One possibility (without variable importance) would be to display partial dependence plots, which show you the connection between the variable and (one) predicted class. See partialPlot in randomForest package in R for more information.
You can also use the Partial prediction options in mlr http://mlr-org.github.io/mlr-tutorial/devel/html/partial_prediction/index.html
